Search strategy

The systematic review was conducted according to the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [20]. A comprehensive search was performed using three electronic medical databases (PubMed, EMBASE, and Cochrane Library) by two independent authors (GA and VI) from January 2020 to April 2020. To achieve maximum sensitivity we combined the search terms “gut microbio* OR gut OR streptococcus OR microbiome” with some terms related to osteoarthritis and inflammation such as “arthritis OR osteoarthritis OR inflammation OR synovial OR synovitis” and typical anatomical landmarks of disease “hip OR knee” as either keywords or Medical Subject Heading (MeSH) terms. The search strategy was repeated during the peer-review process to further check for newly published articles (October 14th, 2021). The reference lists of all included articles, previous reviews on the topic, and top hits from Google Scholar were reviewed to identify further potentially relevant studies, which were assessed using the inclusion and exclusion criteria. To avoid overlap with other ongoing review studies, we searched PROSPERO for any similar reviews.

Selection criteria

The gut microbiome is affected by several factors, including dietary patterns [21], pro- and prebiotics, and drugs (antibiotics, proton-pump inhibitors, non-steroidal anti-inflammatory drugs [22, 23]). Inclusion criteria were (1) both clinical and preclinical studies with any level of evidence (2) published in peer-reviewed journals in English, (3) investigated the role of gut microbiome, with or without dietary interventions, and OA pathogenesis or related-symptoms. We excluded studies in which data was not accessible or missing, or those without an available full-text article. We also excluded duplicates and studies with poor scientific methodology as per risk of bias assessment. Abstracts, case reports, conference presentations, reviews, editorials, and expert opinions were excluded.

Two authors (EC and MW) performed the search and evaluated the articles independently. An experienced researcher in systematic reviews (JP) resolved cases of doubt. 465 articles were retrieved from the three databases after removal of duplicates. First, each of the two investigators read the abstracts of all articles, selected relevant articles according to the inclusion and exclusion criteria, and defined a list of articles for full text reading. Then, the two investigators proceed to full text reading, selected the articles eligible for inclusion and compared the results. An experienced researcher in systematic reviews (JP) resolved cases of doubt. After four weeks, the same studies were reread to ensure the investigators agreed about article selection. There was no disagreement among the investigators. One investigator (EC) extracted data from the full-text articles into an Excel spreadsheet with structured tables to analyze each study descriptively. Doubts and inconsistencies were grouped and resolved. Ultimately, 19 articles were found eligible to be included in the study.

Data extraction and criteria appraisal

Data were extracted from article texts, tables, and figures using the Population, Intervention, Comparison, Outcome framework [24] (PICO) and included the title, year of publication, study design, sample size, study population, patient characteristics, intervention and comparator group (if applicable), outcomes, findings, and conclusions. Two investigators (GA and GG) independently reviewed each article. Discrepancies between the reviewers were resolved by discussion and consensus (Fig 1). After extraction, the data was considered of heterogenous nature both by study design, measure, and method of assessment. Therefore, a descriptive analysis approach was preferred to a metanalysis.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. A PRISMA flowchart showing the selection process.

https://doi.org/10.1371/journal.pone.0261353.g001

Risk of bias assessment

Risk of bias assessment of non-randomized clinical studies was performed according to the ROBINS risk of bias tool [25] for non-randomized study and RoB 2.0 [26] for randomized studies. This assessment used “low,” “moderate,” and “high” as judgment keys. “Low” indicated a low risk of bias, “moderate” indicated a moderate risk of bias, and “high” indicated a high risk of bias. The assessment was performed by two authors (EC and AWF) independently. The inter-rater agreement was 90%. Any discrepancy was solved by consensus. The risk of bias assessment for preclinical studies was instead performed using the SYRCLE’s tool [27]. A score ranging from 0 (lowest) to 10 (highest) was used to classify potential study bias. Two investigators evaluated the studies independently (EC and AWF) with a 95% inter-rater agreement. 12 studies were retrieved after full-text reading and were assessed for the risk of bias. One clinical study was excluded because of a serious risk of bias. Three clinical studies were assessed to have a moderate risk of bias, and one article was excluded because of a serious risk of bias. One randomized clinical study was assessed to have “some concerns” according to the RoB tool. The preclinical animal studies were assessed as having a low to moderate risk of bias (mean SYRCLE score 3.9, range 3–6). (Supplementary Material) (AUTHOR NOTE: weblearn does not allow for supplementary material submission, happy to share by email).

Study quality assessment

The quality of clinical studies was assessed using the GRADE method [28]. Each study was classified as “low”, “moderate” or “high” according to the quality of evidence. With the exception of one study that featured a low quality of evidence (already excluded after the risk of bias assessment), all the studies ranked “high”.

The Collaborative Approach to Meta-Analysis and Review of Animal Data from Experimental Studies (CAMADARES) checklist [29, 30] was used to assess the quality of the preclinical studies (n = 10) (Supplementary Material) (AUTHOR NOTE:weblearn does not allow for supplementary material submission, happy to share by email). The assessment was performed independently by two authors (EC, AWF) with an inter-rater agreement of 94%. Each study was assessed and scored on a scale from 0 (lowest) to 10 (highest) points. The overall quality was moderate among the included studies (mean CAMADARES score 4.45, range 3–6).

Study characteristics

19 studies were included in the final analysis, of which 10 were on animal models [31–38], 8 were human clinical or ex vivo studies [5, 18, 19, 38–47], and one was based on analysis of publicly available databases [40].

Concerning the human clinical trials, 3 were non-randomized observational studies [5, 18, 19], and 3 were randomized clinical trials [41–43]. The main features of the studies included are reported in Table 1. Data was retrieved from 1,548 individuals, of which 58.2% were women. The population was aged between 50 and 65 years. The models investigated in the preclinical studies were adult mice [33, 36, 48–50], and adult rats [31, 34, 35] (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. The main findings of the included studies.

https://doi.org/10.1371/journal.pone.0261353.t001

Most studies analyzed the relationship between OA and the variations in microbial populations of the gastrointestinal tract or in the local articular cartilage (specifically knee and hip samples) by 16S ribosomal RNA gene sequencing, internal transcribed spacer (ITS) amplification and/or shotgun metagenomics [18]. Overall, all studies showed different level of evidence of the role of the gut-joint axis in the pathogenesis and progression of OA.

One study analyzed, in a model of sex steroid deficiency, the role of specific microbial species and strains (administrated by oral gavage of probiotic supplementation) in the maintenance of intestinal barrier integrity. Metabolic alterations of the microbiome and their correlation to gut permeability were investigated by measuring the serum concentration of lipopolysaccharides (LPS), a component of the outer membrane of Gram-negative bacteria [18, 19, 31, 32, 35, 36]. Diet and obesity were analyzed concomitantly [31, 32, 35] in relation to gut microbiome in mice.

Some studies analyzed the role of other dietary aspects in relation to OA severity: physical exercise [31, 49], the administration of nutritional supplementation [34] and/or prebiotics in rodents [31–34] and humans [41] (Table 1).

Effects of diet regimens and the gut-joint axis

High-fat diet (HFD) and steroid-deficiency induced obesity are common models of gut epithelium disruption, gut dysbiosis, and low-grade systemic inflammation [31, 35].

When mice were on an HFD, OA-like and inflammatory changes were observed during histological analysis. However, when pre-and probiotics were administered contextually to the food, the inflammatory effects on the gut-joint axis were partially reversed and prevented [31, 35]. Another study, reported similar results with the administration of oligofructose in obese mice with increased gut permeability [32]. However, Collins et al. [38] reported how these changes can happen even independently of HF diet, but based on direct relationship between the microbiome and the OA phenotype. A study tried to replicate what was previously found by investigating the gut microbiome and LPS level in a case-control study of OA vs non-OA human patients [39]. While alpha e beta diversity did not differ among the two groups, the LPS level were strongly associated with OA severity (104.9 ± 45.8 EU/mL vs 61.3 ± 33.9 EU/mL; p<0.0001) highlighting the role of gut permeability in OA.

Few studies investigated the use of glucosamine sulfate in similar models, and while a non-blinded randomized clinical trial [41] found that both Glucosamine sulfate and Green-lipped mussel used as prebiotic, were effective in improving joint function and pain in knee OA patients. However, a preclinical trial showed how glucosamine was effective only in co-administration with Lactobacillus supplementation [34], which was shown to prevent the changes even when administered alone.

Microbial composition changes associated with osteoarthritis

When obesity was induced by administering a HFD in mice models, the most common finding was an increased Firmicutes/Bacteroidetes (F/B) phyla ratio in the gut compared to chow-fed mice [31, 32, 35] (Table 1) (F/B = 5–10 [35]. This finding was found to be strongly correlated with OA severity according to the modified Mankin Criteria, using a histologic/histochemical grading system.

Quantitatively, the F/B ratio increase has shown to be mostly dependent of a decrease in Bacteroidetes phylum population (especially Bacteroides and Prevotella genera), whereas the total Firmicutes abundance mostly remained the same. However, the Firmicutes composition was different in OA patients, mainly showing a decrease in Lactobacillus spp. and an increase in the Clostridiales order [35]. The largest OA human population investigating the gut microbiome, showed that the relative most abundant species among Lactobacillales, was Streptococcus spp., which have shown to represent 20% of Firmicutes in OA patients and this was positively correlated with higher WOMAC OA pain scores and lower functionality [5]. These results were validated by replication in an independent cohort and by meta-analysis of the final results. Finally, they presented evidence that this association was driven by local inflammation (synovitis and joint effusion) in the joint assessed through a subgroup MRI analysis in 373 patients. Interestingly, a recent study looking a publicly available sequencing data, used mendelian randomization approach to identify microbial taxa associated to OA in humans [40]. Their finding were in agreement with what previously reported.

Gut permeability as a foundational element of the gut-joint axis

Gut microbiome changes such as bacterial abundance, activity or diversity, are observed in animal models of obesity, and can be either acute or chronic. Indeed, obesity is associated with specific microbiome patterns that ultimately lead to impairment of the gut mucosa and microbiome translocation [51]. All of these changes in the gut are commonly described as gut dysbiosis. Interestingly, together with physical exercise, weight loss is strongly associated with OA control [52]. Both interventions are associated with reduced pain and disease progression [53, 54]. In addition, along with functional food use [33, 41], these interventions have been shown to be effective in reverting the acute microbiome compositional changes that are associated with OA, as depicted in this systematic review.

When gut dysbiosis occurs as a result of an underlying chronic disease (e.g., inflammatory bowel disease), chronic antibiotic treatment, or lifestyle modification (e.g., obesity and metabolic disease), gut permeability is significantly affected [55–60]. It has been demonstrated that a HFD depletes eosinophils in the gut and is associated with an increase in gut permeability [61], reduced tight-junctions molecules, mucus, and antimicrobial peptides production [62]. Moreover, microorganisms are spatially redistributed in the intestine, mainly occupying intervillous/cryptal spaces [62]. If such changes progressively occur, innate immune receptors in the gut become activated by microbial products (such as LPS) and stimulate the production of pro-inflammatory mediators (IL-1beta; IL-17, IL-18) both locally and systemically once similar byproducts reach the systemic circulation [39]. Thus, both inflammatory mediators and bacterial toxins (e.g., LPS) are subsequently translocated into the circulation [57] with priming action on the immune system that has been associated with multiple chronic conditions [63–67] including OA [39].

Together with molecular products, bacterial translocation is possible and raises concerns, particularly in patients with prosthetic implants or artificial valves [68, 69]. Although emerging evidence suggests that bacterial translocation is associated with OA severity in the native joint, it is still unclear whether an increased risk for infection of the joints does exist in obese patients or in patients with inflammatory bowel disease. Ultimately, low-grade inflammation of the gut may lead to transient bacteremia, in which either direct colonization of bacteria to the joint may occur or in which leucocytes and macrophage [32, 70] may act as “Trojan Horse” [68, 69, 71] for bacteria to reach the joint. This concept needs to be further explored in subsequent studies.

Adaptive and innate immune response

The aforementioned short-term alterations of the gut microbiome composition are associated with worse patient-reported outcome measures and with a more severe radiologic OA classification [5, 19]. Our findings indicate that this is probably due the observed increase in the Firmicutes/Bacteroidetes (F/B) phyla ratio of the gut microbiome [31, 32, 35].

Bacteroidetes produce high levels of short-chain fatty acids, which regulate the differentiation of Treg cells that are the main actors of the suppression of inflammation in chronic and acute condition [72]. Thus, we propose that the observed clinical findings are due to a decrease in Bacteroidetes, as displayed in OA patients, through the impairment of adaptive immunity and the delicate balance of pro-and anti-inflammatory mechanisms. Also, the abundance of specific strains, such as Streptococcus spp. and other gram-negative strains seems to be critical to prime both local and systemic inflammation through LPS- or other metabolites-induced macrophage activation [5] via common inflammatory pathways such as NFkB or MAPK [73], or by forming active complexes with Lipopolysaccharide-binding-Protein (LBP), laminin-binding protein, and CD14 [19]. Mechanistic evidence was reported with increased CD14 levels (as a surrogate marker for macrophage activation) in the synovium of OA patients compared to healthy controls [74]. Thus, an increase in F/B ratio and Streptococcus spp. prevalence are probably able to dysregulate the delicate balance of pro-inflammatory markers [34]. However, two of the studies we included in our analysis [36, 37], reported that gut microbiome dysbiosis alleviates the progression of osteoarthritis in mice. Additionally, an antibiotic-induced gut microbiota dysbiosis in OA male mice significantly decreased the relative abundance of Bacteroidetes, but for female mice, the relative abundance of Firmicutes was increased significantly compared with that in Con-OA female mice [37]. While this further highlights the pivotal role of the microbiota as whole in the development, progression and symptomatology of OA, also induces contradictory conclusion about bacterial diversity and F/B ratio. While many conclusion can be made, we hypothesize that any change that can lead to impaired gut barrier integrity and the release of byproducts such as LPS into the bloodstream can perhaps induce immune priming and adjuvate the initiation and progression of OA disease.

We also hypothesize that bacterial molecules, that can potentially act as biomarkers, are involved in OA inflammation. We encourage further studies on microbial exotoxins, cationic bacterial peptides, and other metabolites, to assess their role in the pathogenesis of OA, their influence on quorum sensing and inter-and intra-species interactions in biofilm-based musculoskeletal infections such as periprosthetic joint infections [75].

Pain symptomatology and chemokines

Studies show that the gut microbiome is associated with joint pain [76]. The main actors are cartilage degradation products and other damage-associated molecular patterns either through direct neuronal activation of dorsal root ganglia or by indirect neuro-immune signaling acting on immune cells receptors that in turn stimulate neurons amplifying the mechanism. Apart from cytokines and LPS that may partially explain the onset of pain, chemokines are also reported as potential critical players [70]. Indeed, chemokines were consistently increased in murine animal models [32, 49].

Current treatments for OA focuses on pain relief, but do not target the pathogenesis of the disease [52]. Since the gut microbiome is modifiable by several factors (dietary intervention, fecal transplant, and future microbiome-targeted therapeutics), the gut microbiome is a promising target for future treatment strategies.

Limitations

Our findings have several limitations. First of all, despite the consistent data on the association of OA with specific bacterial strains and phyla ratio in the gut microbiome, we caution the reader about the absence of large human studies other than the few reported. To intervene in the gut-joint axis as part of treatment requires a more profound mechanistic understanding of microbiome-host interactions and detailed characterization of the complex community interactions involved. Current studies, despite the overall moderate to high quality of evidence, still do not allow for such conclusions.

Most studies were conducted on animal models, not fully mimicking the complexity of the human microbiome. In addition, the design of all the studies included do not allow for causality conclusions. However, all pieces of evidence analyzed have shown concordance of findings and clinical studies show similar results [5]. Another major limitation of the analyzed studies is the use of different methods to induce obesity in animal models and the heterogeneity of the studied animal models itself. However, all the different animals and the different models used (high fat diet, high sugar diet, sex steroid deficiency) show the same results. Finally, most animal studies only described the alterations of the gut microbiome after specific diet regimens, and by design are not able to describe a clinically relevant microbiome description. Also, despite the studies reported a control group with different diet regiment (e.g. chow-fed), further studies should investigate the alteration of the bacterial strains, and their association with OA, regardless of the diet, maybe using fecal transplantation as validated in other studies [77]. Finally, bacterial genomic plasticity is missing in most of the studies, where under antibiotic—although subinhibitory—pressure, influence of heavy metals from food or drinking water, and nutritional influence, mobile genetic elements drive in-vivo evolutionary events and adaptation processes in the microbiome, its composition and inflammatory activity.